Inspiration
Film and TV productions must research names, brands, titles, and other real-world references before filming or distribution. This process can be slow and expensive, especially for independent filmmakers. We built PERMISSA to make the first research pass faster, clearer, and evidence-based.
What it does
PERMISSA analyzes screenplay elements and organizes potential clearance risks.
It:
- Identifies people, characters, brands, businesses, and titles.
- Researches them using live web sources through Parallel.
- Shows findings with citations, confidence levels, and clear risk statuses.
- Requires human review when evidence is missing or uncertain.
- Produces research support, not legal advice or clearance certification.
How we built it
We built PERMISSA with:
- Gemini and Google agent technology for extraction and orchestration.
- Parallel Search API for live, cited research.
- Next.js and React for the product dashboard.
- NestJS for the API.
- Google Cloud Run for deployment.
- Firebase Authentication and Firestore for identity and data.
- TypeScript and Zod for typed contracts.
- Deterministic policy rules that prevent unsupported findings from being marked as cleared.
Challenges we ran into
The biggest challenge was preventing the AI from presenting uncertain research as legal certainty. We separated AI-assisted research from deterministic evidence rules.
We also worked through secure authentication across multiple Cloud Run services, cost limits, deployment workflows, and strict time constraints.
Accomplishments that we're proud of
- Built a risk-first product instead of a general screenplay chatbot.
- Added citation-backed research and evidence requirements.
- Created deterministic confidence and status rules.
- Added human-review and legal-safety boundaries.
- Deployed separate operator and product experiences.
- Created a synthetic screenplay and expected findings for testing.
What we learned
We learned that responsible AI systems need more than prompts. They need evidence, typed contracts, deterministic rules, cost controls, clear failure states, and human authority.
We also learned that legal-related workflows should support qualified professionals rather than attempt to replace them.
What's next for PERMISSA
Next, we plan to complete durable storage for every research run, connect the full dashboard to live API data, improve screenplay parsing, strengthen testing, and validate the workflow with film-clearance professionals.
In the future, PERMISSA could support title clearance, rights tracking, music and footage permissions, version comparisons, and professional review packets.
Log in or sign up for Devpost to join the conversation.